Share your thoughts, 1 month free Claude Pro on us
See more
Home
/
Benchmarks
Expert-level Question Answering on GPQA-D (Accuracy @8)
Loading...
68.9
Accuracy @8
CERL full
57.668
60.584
63.5
66.416
May 22, 2026
Accuracy @8
Updated 1mo ago
Evaluation Results
Method
Method
Links
Accuracy @8
CERL full
Training=SFT+RL
2026.05
68.9
Halo
Training=SFT+RL
2026.05
66.3
Fixed-ratio erasure + free answer
Training=None
2026.05
66.1
Base Qwen3-14B
Training=N/A
2026.05
64.1
Length-penalty RL
Training=RL
2026.05
63.3
TokenSkip†
Training=SFT
2026.05
58.1
Feedback
Search any
task
Search any
task